Give your firm back the week it loses to documents.
DilloLex is a document engine that runs on hardware your firm owns outright, with software you license once and keep. No monthly seat fees, no per page meter, no vendor sitting between you and your own files. Drop a thousand page client dump in the matter folder, and get back a named, Bates stamped, redaction flagged record in minutes, ready for your team to review. It all happens in your office, and nothing is final until your attorney approves it.
You buy it once and own it. The hardware is yours, and your raw client files are processed on your own network, never sent to an outside AI.
One time license, quoted per firm, plus the appliance you buy direct for about $4,000. One purchase, no per page meter.
Not the attorney? See how DilloLex works for your paralegals and legal assistants →
Jump to: How it works · Proof · Research · ROI · Security
Read access, plus human-approved calendar & task writes.
Your draft tells DilloLex which exhibits to pull.
Building an exhibit set manually means combing the whole record for every document a motion cites, then stamping and indexing each one and praying you missed nothing. DilloLex runs it backward. Hand it the draft you already wrote: it reads every document your motion references, finds each in your OCR'd record, and assembles them into a Bates stamped, indexed set in the order you cite them. You confirm each match. Anything it cannot place with confidence comes back flagged as unmatched, so you see the gap instead of the wrong document.
"...as reflected in the parties' text messages of April 14, 2024 (Ex. A), the March 2024 account statement (Ex. B), and the responding party's deposition transcript (Ex. D)."
Ex. A · Text Messages, Jane to John, Apr 14 2024 matched · Bates 0001-0007
Ex. B · Wells Fargo ····8821 Statement, Mar 2024 matched · Bates 0008-0011
Ex. D · Deposition transcript referenced but missing from the record · needs your input
Illustrative example for display only. DilloLex proposes the matches; your staff confirms each one, and an unmatched reference like Ex. D is flagged rather than guessed. The same reference reading engine compiles the NRAP 30 appendix from your brief and the NRCP 16.2 / 16.205 production index, each Bates stamped with a redaction pass.
Clients hand you "IMG_4471.pdf" and "scan_002.pdf."
Then someone opens each one, figures out it's the 2020 tax return or a text thread between the parties, renames it, and files it. Only then does the slow work begin: building exhibits, running Bates stamps, and marking redactions in Acrobat, one document at a time.
Hours of manual triage
Opening, reading, and renaming unintelligible files is costly, low value work that buries your staff.
Exhibit & disclosure churn
Assembling NRCP 16.2/16.205 disclosures, PI demand packages, and Bates stamped exhibits takes days.
Confidentiality pressure
Marking SSNs and PHI for redaction page by page in Acrobat is slow and easy to get wrong. Many AI tools also send your client files to the cloud.
Watch it make sense of a mystery file.
Tap a junk filename and see what DilloLex would call it.
This runs today. The names are configured for your firm during onboarding. DilloLex names the bulk on its own, and your staff steps in only on the few it cannot place with confidence.
Illustrative example for display only.
The engine is live. The appliance puts it in your office.
The document engine runs on real Nevada matters today, across family law, personal injury, and appellate. DilloLex is delivered as an appliance: dedicated hardware that sits in your office, runs the AI locally, and keeps your files on your own network.
In the engine today
One pipeline, running locally, inside the storage folders your staff already use. No uploads, no new app.
Onboarded to your firm
The appliance runs in your office on hardware your firm owns. It is onboarded to your firm's templates and document conventions. It includes Litigation Intelligence: before you draft, you see how the judge on your matter has ruled and which arguments opposing counsel recycles. Every finding cites the source document and page. The attorney reads it and decides.
Exhibit assembly from your draft
Reads every document your motion references, matches each one in your matter record, and builds a Bates stamped, indexed exhibit set; unmatched references come back flagged for you to resolve. The same engine compiles NRCP 16.2 / 16.205 disclosures and PI records packages.
PII/PHI detection and redaction
Flags SSNs, account numbers, and PHI, then burns the redactions you authorize, with a QA pass that flags likely misses. Final verification stays with your staff.
Matter organization and search
OCRs every page (English and Spanish), proposes a clean name for each file, and lets you search the whole record in plain language with cited source passages.
Template drafting and deadline flagging
Drafts reviewed by attorneys, built from your own templates, plus rule based Nevada deadlines read from a filed order for your docketing to verify.
Litigation Intelligence
Before you write a word, see what actually moves the judge on your matter: which facts, exhibits, and arguments win, and which mistakes get opinions reversed. See the arguments and authorities your opponent recycles, so your opening motion answers them first. Every finding comes from our own corpus of appellate decisions and cites the source document and page. The attorney reads it and decides. The memo informs your strategy; it does not replace your judgment.
A thousand pages, searchable in about three minutes.
OCR a 1,000 page scan in Adobe Acrobat and your paralegal loses most of an hour, stuck at a machine grinding through one page at a time. DilloLex runs the whole job in parallel on a dedicated machine, in the background, and it does not stop at OCR: it names, Bates stamps, flags redactions, and files.
| 1,000 scanned pages | Paralegal's PC + Adobe Acrobat | DilloLex |
|---|---|---|
| Time to searchable text | ~35 to 65 minutes | ~3 minutes |
| How it runs | One page at a time (serial) | In parallel, on a dedicated machine |
| The computer doing it | Tied up; the paralegal waits | Free; runs in the background |
| When the OCR finishes | Staff still name, number, redact, and file | Named, Bates stamped, flagged, queued for review |
Measured on our document engine, a 16 core machine: a 1,000 page, 300 DPI document OCR'd in parallel to a searchable text layer with Tesseract via OCRmyPDF. The Adobe figure is a commonly cited single threaded OCR rate on a typical workstation (about 2 to 4 seconds per page). Real times vary with scan quality and hardware; the operational point is that the appliance does it unattended and in parallel while your staff's PC stays free.
Three real Nevada appeals. Measured on actual matters.
Supervised test runs, with the attorneys in the loop on real matters. What the engine produced:
Under 3 hours
Opening brief (~12,000 words) and its full NRAP 30 appendix, start to finish.
Under 30 minutes of attorney revisions.
Under 3 hours
A second appeal: opening brief (~8,000 words) plus its appendix.
Under 30 minutes of attorney revisions.
Under 45 minutes
A respondent side answering brief (~10,500 words), the other side of the caption.
Only minor revisions.
All three came in under the NRAP 14,000 word limit, a factual count, and on attorney review each complied with the applicable appellate rules.
First results on a small sample, and your results may differ. Every output, including every legal argument, was reviewed and approved by an attorney before use. The engine produced these on real Nevada matters; the appliance that delivers it in your office runs that same engine locally. No client, firm, or case names are disclosed.
Search the law, not just your files. Every answer is grounded in cited authority.
DilloLex builds its research on a local corpus of Nevada appellate decisions, statutes, court rules, and secondary sources. You get a cited answer with the source passage, not a confident guess from an outside model that has never read Nevada law in any systematic way.
Nevada first, built from primary sources
The engine indexes Nevada Supreme Court and Court of Appeals opinions, NRS statutes, court rules, and administrative regulations. Every answer cites the document and the passage. If the corpus does not contain a controlling source, the engine says so rather than filling the gap with a fabricated citation.
Citation grounding by design
Hallucination is the central failure mode in legal AI: a system that fabricates cases and quotations is worse than no system at all. Every research answer DilloLex produces is retrieved from the corpus and paired with the source document, so the attorney can read the passage before relying on it. The engine proposes; the attorney verifies.
Honestly scoped to what it knows
The research engine is built around Nevada law, and it is calibrated to know its own limits. It labels what the corpus contains, shows you the source for every answer, and flags when a question falls outside what it covers instead of guessing. You always know the ground an answer stands on.
It rarely invents a citation.
In law, one number matters most: how often a tool makes up a citation. On about 200 Nevada questions with known answers, ours came in at about 5 in 100. Stanford put the leading paid tools far higher: about 17 in 100 for Lexis and 33 for Westlaw, on their own separate tests. This was our own internal test, not a side by side test. See how we keep our AI honest →
Our Nevada test, about 200 questions
Every question in the benchmark set is a Nevada legal question with a verifiable answer drawn from primary authority. We tested across statutory interpretation, procedural rules, case law, and deadline questions. An answer is scored correct only when grounded in cited authority that actually says what the answer claims. Methodology is documented and available to firms and carriers on request.
Methodology documented. Available to firms and carriers on request.
Every answer is auditable by the attorney
The accuracy story matters only if you can verify it yourself. Each research result shows the source passage alongside the answer, in your own corpus, on your own machine. You do not have to trust our benchmark: you can read the source the engine is citing and decide whether it supports the answer.
Source passages returned with every answer. Fabricated citations have no passage to show.
More billable work. No new hires.
Your team turns out the same briefs and motions in a fraction of the hours. The hours they get back are hours you can bill, which means more revenue without hiring. See the full savings model.
Illustrative. Freed hours become revenue only if you fill them with billable work, and attorney review still takes time. In our testing, an opening brief and its appendix went from days to under 3 hours. See if it pays for itself.
Six questions worth asking any AI vendor.
Whether a tool runs in the cloud or in your office, the same questions separate a confidentiality safe choice from a leap of faith. Ask every vendor these before a client file goes anywhere.
- Where does my client's data physically go when I use this?
- Is my data, or anything derived from it, used to train or improve a model?
- Is it retained, for how long, and can I delete it on demand?
- Can a human at the vendor read it, and under what circumstances?
- What do the actual contract terms say, as opposed to the marketing page?
- Whose hardware runs the AI, mine or the vendor's, and is any outside AI service in the path at all?
Your documents stay armored in your office.
Most legal AI tools ship your client files to someone else's cloud. DilloLex is built the other way: the AI runs on the appliance your firm owns, and no client or confidential content ever reaches an outside AI service.
Stays on your network
No inbound internet access. Encrypted disk. Raw files processed locally.
The AI runs on the box
Drafting and analysis are built to run on the appliance your firm owns, right inside your office. No outside AI service ever sees your client content.
No AI account to keep
No API key, no per token bill, no outside model in the loop. The reasoning lives on hardware your firm owns.
By design, all AI reasoning runs on the appliance your firm owns, so no client or confidential content is ever sent to any outside AI service. The box still connects to your own cloud storage and the tools you choose, so it is not sealed off from the internet. The scope is that no outside AI service is in the loop.
The tools built cloud first send your client files out. DilloLex keeps them in.
The current wave of legal AI (Supio, EvenUp, Eve, Spellbook) runs in the cloud. Your client files go to a vendor's servers. DilloLex runs on hardware you own, built for firms that won't make that trade.
| DilloLex · in your office | Cloud legal AI | Manual / Acrobat | |
|---|---|---|---|
| Where your files live | Your office | Vendor cloud | Your office |
| Who holds the data | You | The vendor | You |
| Per document fees | None from DilloLex | Often priced by usage | No (flat license) |
| Works without sending files out | Yes | No | Yes |
| You own the hardware | Yes | No | Yes |
| Output looks like your firm's filings | Yes | Generic template | Yes |
| Speed without files leaving the office | Minutes, nothing uploaded | Minutes, after upload | Hours |
The document spine works across every practice area.
OCR, classify, Bates stamp, redact, assemble exhibits, and compile appellate appendices. The pipeline works across every practice area. What changes is which filings you build. A few examples:
Family Law
Organize the records behind a divorce or custody matter, build the NRCP 16.2 / 16.205 disclosure compilation with a Bates ready index, and assemble hearing exhibits.
Personal Injury
From intake through litigation: tabulate medical records and bills, assemble the indexed demand package, and build early case disclosures and trial exhibit sets, so your team spends its time building the file.
Appellate
Compile the NRAP 30 appendix from the record into indexed volumes that conform to the rules. Read a brief and pull the authority compendium. Appellate document work runs on the same pipeline as every other matter.
Criminal, Probate & General Civil
Organize the client file, Bates stamp and exhibit any record, run redaction QA on produced documents, and build trial exhibit binders, regardless of practice type. If your firm handles it in Nevada, the document pipeline handles it too.
It still looks like your work.
You spent years building how your firm's filings look. Generic AI tools flatten all of that into a single template. DilloLex loads your format during onboarding and drafts from your own templates. The finished product still looks like your work.
Your pleading paper
It learns your filing format from your own samples (the page margins, the 1-to-28 line numbering, the borders, the header and footer, the attorney and firm block) and places every filing on it.
Your font and layout
The base typeface and page layout your firm has always used, carried onto every document, so a finished draft reads as yours and not as something a tool generated.
Your templates and language
Demand packages, letters, and disclosures are drafted from your firm's own templates and standard language, for your attorney to finalize, every one grounded in your own forms.
Built by an attorney who ran these workflows.
DilloLex is jointly owned by an attorney licensed in Nevada and Washington who ran these exact workflows in active practice. We document the exact data flow with you in writing before you launch. We will walk your malpractice carrier through it.
DilloLex runs in your office, on hardware your firm owns. No outside AI sees your files, no monthly seat fees, no per page meter. You buy it once.
Bring DilloLex to your Nevada firm.
DilloLex runs on real matters in your office today.
Book a demoNot ready to talk? Read the two minute RPC 1.6 security brief →
Runs in your office. You own the hardware and the data. No obligation.